Notes on a life with sound.

Voice is a vital sign — but voice is also one instance of a larger idea. Acoustic signal, continuously and naturally emitted by living things, carries far more information about their states than the world has learned to use. My work has been to make that information usable.

My professional career has moved in that direction along three paths. In bioacoustics, the Owl Project at the MIT Media Lab established that meaning can be inferred from sparse, ambient, uncontrolled acoustic samples gathered over commodity infrastructure — cell phones in forests, listening for owls and inviting them to answer. In signal-processing theory, work at Tulane, Michigan State, and MIT Lincoln Laboratory developed the algorithmic foundations that make such inference possible: set-membership identification, optimal bounding ellipsoid methods, vocal-tract coordination features, the careful mathematics of listening under uncertainty. In clinical voice AI at industrial scale, the teams I led at Sonde Health partnered with clinicians on multi-continent longitudinal studies in respiratory and mental health, reached on-chip implementation, and confronted the questions that now define the field — adoption, privacy, dataset diversity, the gap between cross-sectional data and clinical decision-making.

Across those three communities, the underlying problem has been the same: how to make a passive, ambient, naturally occurring acoustic signal carry weight as evidence. The technical layer of that problem — triggered sampling, in-the-wild quality, trajectory-native models, inference at the silicon edge, privacy architecture — is hard. The human layer is harder. Trust, adoption, the patience to do this slowly and well. Voice health in particular sits at the intersection of all of it, and the populations with most to gain are the cohorts least well served by today’s appointment-based, hospital-anchored system: the elderly, whose voices change as their bodies do; those living in under-resourced regions, whose voices have largely been absent from the datasets on which the field has been built; and patients whose care has historically been least continuous.

This website also contains personal threads, because they inform the work. Music for one. Some of the wider thinking shared through a Forum column for The Carlisle Mosquito. K–12 technology workshops which translated current work in acoustics, sensing, and computing into something students can hold in their hands. Outreach as a structural commitment rather than a footnote, because the same ear that learned to listen to owls and to patients learned much of what it knows by being listened to first.

See also: selected writings on the Writings page, project chapters on Practice, and the K–12 workshop work on Outreach. Complete CV available on request.